Repeater Production for the
North Atlantic Link
By H. A. LAMB* and W. W. HEFFNER*
(Manuscript recrived September 20, 1956)
Production of submarine telephone cable repeaters, designed to have a
minimum trouble-free life of iwenlij years, required many new and refined
manufacturing procedures. Care in the selection and training of personnel,
manufacturing environment, inspection, and testing, were of great impor-
tance in the successful attainment of the ultimate objective. Although quality
of product has always been of major significance in Western Electric
Company manufacture, building electronic equipment for use at the bottom
of the ocean, where maintenance is impossible and replacement of apparatus
extremely expensive, required unusual manufacturing methods.
Late in 1952, the manufacture of flexible repeaters for the North
Atlantic Link of the transatlantic submarine telephone cable system
was allocated to the Kearny Works of Western Electric Company.
In accordance with established practice in initiating radically new
products and processes, production of these repeaters was assigned to
the Engineer of Maiuifacture Organization rather than to regular manu-
facture in the telephone apparatus shops. The job — to produce 122
thirty-six channel carrier repeaters and 19 equalizei-s capalile of operat-
ing satisfactorily at pressui'es up to 6,800 pounds per square inch on the
ocean floor, with minimum maintenance, for a period of at least twenty
years. Initial delivery of repeaters was reciuired for March, 1954, less
than a year and a half after the project started.
Quality has always been the piime consideration in producing appara-
tus and equipment for the Bell System. There is an economical breaking
point, however, beyond which the return does not warrant the abnormal
• Western Electric Company.
lO-t THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957
expenditures rc(iuircd to approach theoretical perfection. The same
philosophy applies to all manufactured commodities, be they auto-
mobiles, airplanes or telephone systems. In general, all of these products
are physically available for preventive and corrective maintenance at
nominal cost. With electronic repeaters at the bottom of the ocean, main-
tenance is impo.ssible and replacement would be extremely expensive.
The general philosophy adopted at the inception of the project was
to build integrity into the product to the limit of practicabihty. To do
this, a number of fundamental premises were established, which form
the foundation of all operations involved:
1. Manufacturing environment would be provided which, in addition
to furnishing a desirable place to work, could be kept scrupulously clean
and free from contamination.
2. The best available talent would be screened and selected for the
particular work involved.
3. Wage payments would be based on day work, rather than on an
incentive plan basis, because production schedules and the complexity
of the operations did not permit the high degree of standardization
essential to effective wage incentive operation.
4. A sense of individual responsibility would be inculcated in every
person on the job.
5. Training programs would be established to thoroughly prepare
supervisors, operators, and inspectors for their respective assignments
before doing any work on the project.
(i. Inspection, on a 100 per cent basis, would be established at every
point in the process which could, conceivably, contribute to, or affect
the integrity of the product.
PREPARATION FOR MANUFACTURE
It appeared desirable to set up manufacture in a location apart from
the general manufacturing area. Experience gained to date has satisfied
us that this was the correct approach, since it provided a number of
1. Administration has been greatly facilitated by having all necessary
levels of supervision located in the immediate vicinity of the work.
2. It was necessary for the people on the job to acquire and maintain
a new philosophy of perfection in product, rather than a high output
at an "acceptable quality level." This was easier at a separate location,
since only one philosophy was followed throughout the plant.
FLEXIBLE HEPEATEK AL-iNUFACTUEE
3. Engineering, production control, service and maintenance organi-
zations were located close to actual production and had no assignments
other than the project.
4. The small plant, due to its semi -isolation, tends to produce a very
closely knit organization and good teamwork.
A large number of manufacturing locations were examined and the
one selected was a one-story modern structure in Hillside, New Jersey,
which provided a gross area of 43,700 square feet.
The entire plant was air conditioned ; in most cases, the temperature was
controlled to minimum 73 degrees F, maximum 77 degrees F. The air was
filtered through two mechanical and one electrostatic filters. Relative
humidity was maintained at maximum 40 per cent in all but one area ■ —
the capacitor winding room — in which it was necessary to maintain
maximum 20 per cent humidity to avoid mechanical difficulty with
capacitor paper. While mo.st of the air was recirculated, the air from the
cafeteria, cleaning room, locker and toilet rooms was exhausted to the
CLASSIFICATION OF AREA
Fig. 1 ^ Plant layout.
106 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957
outside atmosphere. Two separate air conditioning systems were in use.
One, of 300 tons capacity, provided for most of the plant, while a smaller
unit of 30 tons capacity served the capacitor winding, testing, and im-
pregnating rooms. Each installation had its own air filtering and condi-
The plant layout is illustrated in Fig. 1. All working areas, with the
exception of the repeater enclosure area, were individually enclosed,
and walls from approximately four feet above the floor were almost
entirely of reinforced glass. This arrangment facilitated supervision by
other than first-line supervisors, who were located with the groups, and
provided a means of viewing the operations by the many visitors at
Hillside, without contaminating the critical areas or disturbing the
Analysis of Design for Facilities and Operations
In analyzing the design for manufacture there were, of course, numer-
ous instances where conventional methods and facilities were entirely
adequate for the job. Since their inclusion would contribute little to this
article, we shall confine the description to those cases which are new or
Collaboration with Bell Telephone Laboratories in Preparation of Manu-
f adoring Information
Early in 1953 a coordination committee was established, consisting
of representatives from the various Laboratories design groups and
Western engineers, which met on a bi-weekly basis during the entire
period preceding initial manufacturing operations. These meetings
provided a clearing house for questions and policies of a general nature
for this particular project and served to keep all concerned informed as
to the progress of design and the preparations for manufacture.
It is customary, during the latter stages of development of any project
at the Laboratories, for Western engineers to participate in the prepara-
tion of manufacturing information as an aid in pointing the design to-
ward the most economical and satisfactory production methods and
facilities. Since the decision to use the Bell System repeater in the Trans-
atlantic system was based on the performance of the Key West-Havana
installation, and the fact that changes in design would require further
FLEXIBLE REPEATER MAXUFACTURE 107
trials over an extended period of time, only minor changes to facilitate
manufacture were made. Further, since some experience had been gained
by the Laboratories in producing repeaters for that installation, it was
decided to "pool" effort in preparing the manufacturing process informa-
tion, which is normally Western's responsibility. Close cooperation of
the two groups, therefore, has resulted in the production of repeaters
wliich are essentially replicas of those in the initial installation except
for the internal changes necessary to increase transmission capacity
from 24 to 36 channels.
Other Western Electric Locations and Outside Suppliers
During the development work on the Key West-Havana repeaters,
the Hawthorne Works of Western Electric had furnished the molyb-
denum-permalloy cores for certain inductors, the Tonawanda Plant
had furnished mantlrelated resistance wire, and the AUentown Plant
had fabricated the glass seal subassemblies. Since the experience gained
in this development work was extremely "\'aluable in producing the
additional material required for the Transatlantic system and since
the facihties for doing the work were largely available, these various
locations were asked to furnish similar material for the project. Although
the Kearny Crystal Shop had not been involved in the Key West-Havana
project, arrangements were made there to make the crystals for this
project, since facilities were available, along with considerable experience
in producing precision units.
While it was believed, initially, that all component parts for repeaters
should be manufactured by Western Electric, critical analysis indicated
that it was neither desirable nor economical in certain cases. One of the
outstanding examples in this category is the hardened and ground
chrome-molybdenum steel rings that constitute the strength members
in the repeater and sustain the pressures developed on the ocean bottom.
Purchasing the many large and varied machine tools and associated
lieat treating eciuipment necessary to produce these parts would have
re(iuired a substantial capital expenditure and additional manufacturing
space. Arrangements, therefore, were made with a highly qualified and
well equipped supplier to produce the rings, using material furnished by
Western, which liad been pre\-iously inspected and tested to very strin-
lOS THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957
The situation attending the manufacture of a relatively small number
of comparatively large copper parts used in the rubber and core tube
seals was much the same. Here, again, the large size machine tools and
additional manufacturing space, required for only a short time, would
have increased the o^'er-all cost of the project considerably. These parts,
therefore, were subcontracted in the local area and inspection was per-
formed by Hillside inspectors.
A safeguard, in so far as integrity is concerned, was provided by the
fact that these were individual parts that could be reinspected at the
time of delivery. No subassembly operations that might possibly result
in oversight of a defect, were subcontracted.
Two major problems confronted us in planning the manufacture of
repeaters. First, to produce units that were essentially perfect; and
second, to prevent the contamination of the product by any substance
that might degrade its performance over a long period of time. In ap-
proaching both of these objectives, it was realized that the product had
a definite economic value which the cost of production should not
exceed. In many cases, therefore, it was necessary to rely on judgment,
backed by considerable manufacturing experience, in determining when
the "point of no return" had been reached in refining processes and
The initial approach to this phase of the job was to classify, with the
collaboration of Bell Telephone Laboratories, all of the manufacturing
operations involved as to the degree of cleanliness required. In setting
up these criteria, it was necessary to evahiate the importance of contami-
nation in each area and the practicability of eliminating it at the source
or to insure that whatever foreign material accumulated on the product
A representative case is the machining of piece parts. While the shop
area is cleaner, perhaps, than any similar area in industry, the very
nature of the work is such that immediate contamination cannot be
avoided since material is being removed in the form of chips and turn-
ings, and a water soluble oil is used as a coolant. In this instance, however,
the parts can be thoroughly cleaned end their condition observed liefore
leaving the area. Conversely, in the case of an operation such as the
assembly of paper capacitors into a container which is then hermetically
sealed, it is vitally necessary to insure that both the manufacturing
FLEXIBLE KEPEATER MANUFACTURE 109
area and the processes are free from, P.nd not conducive to producing,
particles of material which are capable of causing trouble.
The various classifications established for the production areas include
specific requirements as to temperature, relative humidity, static pres-
sure with respect to adjacent areas, cleanliness in terms of restrictions
on smoking and the use of cosmetics and food, and the type and use of
Employees' clothing was considered one of the most important sources
of contamination for two reasons; first, for the foreign material that
could be collected upon it and carried into the manufacturing areas, and
second, that various types of textiles in popular use are subject to con-
siderable raveling and fraying.
After considerable study of many types of clothing for use in critical
areas, the material adopted was closely woven Orion, which has proved
to be acceptably hnt-free. The complete uniform — supplied at no cost
to employees — consists of slacks and shirts for both male and female
employees, Orion surgeon's caps for the men and nylon-visored caps
for the women, hi addition shoes, without toecap seams, were provided.
Nylon smof^ks were furiii.shed to protect the uniforms while employees
moved from locker rooms to the entrance vestibule. Two changes of
clothing were provided each week, and the laundering was done by an
Employees to whom this special clothing was issued were paired for
locker use. Both kept their uniforms and special shoes in one locker and
their own clothes and shoes in the other. This prevented the transfer
to the uniforms of any foreign material that might exist on the street
clothing. At the entrance vestibule to the A, B, and C areas (Fig. 1)
the employees were required to clean their shoes in the specially designed
facilities provided and to wash their hands in the wash basins installed
for this purpose. Smocks were then removed and hung on numbered
hooks that line the walls at the end of the vestibule. Employees were
then permitted to go to their work positions within the inner areas. At
any time that it was necessary for employees to leave the work areas
for any purpose, they were required to put on their smocks in the vesti-
bule and upon their return, to go through the cleaning procedure again.
Employees in the other areas were provided only w^th smocks, mainly
for the protection of their clothes since the work involved could soil or
stain them but could not be contaminated from the clothing.
110 THE BELL SYSTEM TECHNICAL JOUENAL, JANTJAEY 1957
Schedules were established for cleaning the areas at regular intervals,
the frequency and methods depending upon the type of manufacturing
operations and the activity. Usually, the vinyl plastic floors were ma-
chine scrubbed and vacuum dried. Walls, windows and ceihngs were
cleaned by hand with hnt-free cloths. Manufacturing facihties such as
bench tops, which were linoleum covered, were washed daily. Test sets,
cabinets, test chambers and bench fixtures were also cleaned daily.
Hand tools were cleaned at least once a week by scrubbing with a solu-
tion of green soap, rinsing in distilled water, followed by alcohol and
then dried in an oven.
Since it was impossible to determine what contaminating material
in the form of air-borne particles might be encountered from day to day,
and what the effect might be during the life of the repeaters, the general
approach to this problem was to control, so far as possible, the amount
of dust within the plant.
In order to verify, continuously, the over-all effectiveness of the vari-
ous preventive measures, dust counts were made in each classified area
at daily intervals, using a Bausch and Lomb Dust Counter. This device
combines, in one instrument, air-sampling means and a particle-counting
microscope. Over a two-year period it has been possible to maintain, in
certain areas, a maximum dust count of between 2,000 and 3,500 particles
per cubic foot of air with a maximum size of 10 microns. Control checks,
taken outside the building at the employees' entrance, generally run
upwards of 25,000 particles per cubic foot, a good portion of which are
of comparatively large size.
PRODUCTION" AND PERSONNEL
Equipping the plant, obtaining and installing facilities, and selecting
and training personnel proceeded on a closely overlapped basis with
receipt and analysis of Bell Telephone Laboratories' product design
information. Because of the critical nature of the product, provisions
were made not only for the most reliable commercially available utilities
and services, but also for emergency lighting service in some areas.
Maintenance and service staffs had to be built up rapidly as the super-
visory and manufacturing forces were being developed.
FLEXIBLE REPEATER M.\JS'UFACTUHE 111
"Qualification" of All Personnel
Before employecM were assigned to production work they were re-
quired to pass a qualification test established by the inspection organi-
zation to denionsli'ate satisfactory performance. Programs were, there-
fore, set up for "vestibule" training and qualification of new employees.
This activity was carried on by full-time instructors who had been trained
by Western and Bell Laboratories engineers. Training was carried out
in two stages:
1. (a) The employee received instruction and became acquainted
with equipmeut and lequirements. (b) A practice period in which the
employee developed techniques aiid worked under actual operating
(conditions, with all work submitted to regular inspection.
2. A qualification period in which the employee was recjuired to
demonstrate that work satisfactory for project use could be produced.
The main objective during stage 1 was progressive quality improve-
ment and in stage 2 the maintenance of a satisfactory quality'' level over
an extended period of time. Employees made a definite number of units
at acceptable quality levels in order to qualify. The number of units
required for training varied with the type of work and the ease with
which it was mastered.
All personnel were required to pass qualification tests before being
assigned to production work anfl were restricted to that work unless
trained and qualified for other work. Employees trained on more than
one job were requalified before being returned to a previous assignment.
Records of the performance of individual operators started in the
training stage wore continued after the employees were assigned to
production work. The performance record of the operators was based on
results obtained during the inspection of their work, while that of the
inspectors was liased on special quahty accuracy checks of their work.
It was apparent that the new manufacturing techniques, including
the cleanliness and quality demands, would necessitate that all shop
supervisors and employees be very carefully selected. It also appeared
(and this was subsequently confirmed) that after the careful selection
and training of supervisors, long training periods would be required for
specialty selected shop employees.
In selecting first line shop supervisors, such factors as adaptability,
personality, and ability to work closely with the engineers were of para-
mount importance. For the parts and apparatus included in their re-
112 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957
sponsibility, they were required to thoroughly learn the design, the
operations to be performed, the faciUties to be used, the data to be
recorded, the cleanliness practices to be observed — and in most cases,
prepare themselves to be able to do practically all of the operations, be-
cause subsequently they had to train selected operators to perform criti-
cal operations to very high quality standards under rigidly controlled
manufacturing conditions. As shop supervisors and employees were
assigned to the manufacture of repeaters, they were thoroughly indoc-
trinated in the design intent and the new philosophy of manufacture.
Standard ability and adaptabihty tests were used in a large number
of cases to assist in proper selection and placement of technicians. Tests
for finger and hand dexterity; sustained attention; eyes, including per-
ception and observation ; and reaction time of the right foot after a visual
stimulus. (The latter test was relatively important for induction brazing
operations.) Other requisite considerations were a high degree of de-
pendability and integrity, involving intellectual honesty and conscien-
tious convictions; capability of performing tedious, frustrating, and
exasperating operations against ultra-high quality standards, verifying
their own work; perseverance and capability to easily adapt to changes
in assignment and occupation or the introduction of design changes.
We considered whether or not they would stand up under "fishbowl"
operations, wherein they would receive a considerable amount of ob-
servation from high levels of Western Electric Company and Kell Sys-
tem management and other visitors. Also, could they duplicate high
quality frequently after qualifying for a particular operation?
During the period of repeater manufacture, the number of employees
rose from less than 50 in January, 1954, to a maximum of 304 by Feb-
ruary, 1955, after which there was a gradual reduction to a level of about
205 employees for six months and then a gradual falling off as we were
completing the last of the project. In the period from May to December,
1954, between 30 and 45 employees were constantly in training prior to
being placed on productive work. During 1955 this decreased to prac-
tically no employees in training during the midpart of the year and there-
after training was required merely to compensate for a small labor turn-
over and employee reassigim.ient. It is significant that labor turnover
was very low and attendance was exceptionally good during the life of
the Hillside operations.
The original plan, which was generally followed, was to prove in the
tools for each phase of the job, followed by an intensive program of train-
FLEXIBLE REPEATER MANUFACTURE 113
ing. Iiuloctriiuitiou of laboratory technicians could be considered as
"vestibule triiining" in that they were acclimated to the area and con-
ditidUH, given oral instruction in the work, then given practice materials
and demonstrations and, when cjualified, were started on making project
material. To do this, extra supervisors were required at the beginning
of the job. A super\'isor trained a few employees, qualified some of them,
and began work on the project. Another supor\isor was then required
to train additional employees who, as they became qualified, were trans-
ferred to the supervisor re.sponsible for making project apparatus. Addi-
tional testing of the employees, instruction and reinstruction and, in
some cases, retraining were required. In practically all cases, we were
able to fit an employee selected for work at Hillside into some particular
group of operations. The extra emphasis on selection and training cre-
ated a well-balanced team that later resulted in considerable flexibility.
During all of this training our supervisors worked closely with engineers
and inspectors who understood the design intent and the degree of per-
At the bogimiing, each technician was trained for only one operation
of a particular job, such as (1) winding Type X capacitors or (2) im-
pregnating all paper capacitors or (3) winding Type Y transformers and
so became an expert on this one operation. Later, the tours of duty for
many technicians were broadened to cover several operations.
To keep employees informed, wc occasionally assembled the entire
group, presenting informati\'e talks on current production plans and
our future l)usiness prospects. Motion pictures were shown of the cable
laying ships and the operations of cable sphcing and cable laying. A dis-
play l)oard, showing all of the repeater components, was mounted on
the wall of the cafeteria. This informed the operators just where the
parts were used in apparatus; also, just where their products went into
the wired repeater unit, and how all electrical apparatus was enclosed
against sea pressure in the final repeater. In small groups, all of the em-
ployees at Hillside were given a short guided tour of the plant to see the
facilities and hear a de.scription of the operations being performed in
each area. These communications were extended to everyone at the
Hillside Plant, iticluding those who did not work directly on the product.
It was our conviction that the maintenance men, boiler operators, oilers,
station wagon chauffeur, janitors, and clerical workers in the office were
all interested and could do a better job if kept informed of the needs and
progress of the project.
114 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957
Capacity was provided at the Hillside Shop to manufacture a max-
imum of 14 repeaters in a calendar month. This envisioned 6-day opera-
tion with some second and third shift operations; due allowance was
made for holidays and vacations, so that the annual rate would be ap-
proximately 160 enclosures per year. (An enclosure is either a repeater
or an equalizer.)
Some of the facilities and raw materials were ordered late in 1953.
This ordering expanded early in 1954 and continued through 1955 to
include parts to be made by outside suppliers and the parts and appara-
tus to be made at Hillside. Apparatus designs were not all available at
the beginning of the job, and the ultimate quantities required were also
subject to sharp change as the project shaped up, thus further compli-
cating the scheduling problem.
Because of the time and economic factors involved, coupled with the
developmental nature of the product and processes, one of the most
difficult and continuing problems was the balancing of production to
meet schedules. For this task, we devised "tree charts" for the apparatus
codes and time intervals in each type of repeater or equalizer for each
project. Each chart was established from estimates of the time required
to accomplish the specified operations and the percentage of good prod-
uct each major group of operations was expected to produce.
Many of the specifications were written around the specific needs of
the job and embodied requirements that were considerably more strin-
gent than those imposed on similar materials for commercial use. As a
result, it was necessary for many supphers to refine their processes, and,
in some eases, to produce the material on a laboratory basis.
One example is the container, or repeater enclosure, which consists,
in part, of a seamless copper tube approximately If inches in diameter
having a ^-inch wall and approximately 8 feet long. This material was
purchased in standard lengths of 10 feet. The basic material was re-
quired to be phosphorous deoxidized copper of 99.80 per cent purity.
The tubing, as delivered, had to be smooth, bright, and free from dirt,
grease, oxides (or other inclusions including copper chips), scale, voids,
laps, and shvers. Dents, pits, scratches, and other mechanical defects
could not be greater than 0.003 inch in depth. The tubing had to be
concentric within 0.002 inch and the curvature in a 10-foot length not
exceed ^ inch to facilitate assembly over the steel rings.
FLEXIBLE REPEATER MANUFACTURE 115
Only one supplier was willing to accept orders for the tubes, and only
on the basis of meeting the mechanical requirements on the outside sur-
face. To establish a source of supply, it was necessary to accept the sup-
plier's proposal on the basis that some of the tubes produced could be
expected to meet requirements on the inside as well as the outside sur-
face. Inspection of the inside surface was performed with a 10-foot Bore-
The suppher then set aside, overhauled, and cleaned a complete group
of drawing facilities for this project. In addition, a number of refinements
were made in lul)rication and systematic maintenance of tools. After all
refinements were made and precautions taken, however, the yield of
good tubes in the first 400 produced was less than 1.0 per cent. Consulta-
tions with Western and Bell Laboratories' engineers, and with the sup-
plier's cooperation, raised the yield to approximately 50 per cent.
Procurement of satisfactory mica laminations for capacitors intro-
duced an unusual problem. The best grade of mica available in the world
market was purchased which the supplier, under special plant condi-
tions, spht and processed into laminations. Despite care in selection and
processing, only 50 per cent of the 250,000 laminations purchased met
the extremely rigid requirements for microscopic inclusions and delam-
inations, and less than 8 per cent survived the capacitor manufacturing
A large number of the parts, and the most complex, are made from
methyl-methacrylate (Plexiglass). At the time manufacture began, there
was little, if any, experience or information a^'ailable on machining this
material to the required close tolerances and surface finish. Consequently,
considerable pioneering effort was expended in this field before satis-
factory results were obtained.
The methacrylate parts cover a wide range of size and complexity — •
from l^-inch diameter by 4|-inch long tubular housing to tiny spools
|-inch diameter and ■^-inch long. Most of the parts are cyhndrical in
shape with some semi cylindrical sections that must mate with other
sections to form complete cylinders. Others have thin fins, walls, flanges
and projections. Five representative parts are shown in Fig. 2.
Methyl-methacrylate has a tendency to chip if tools are not kept sharp
and care is not used in entry or exit of the tool in the work, particularly
in milling. In some cases, it is necessary, with end-milling, to work the
cutter around the periphery of the area for a slight depth so that sub-
sequent cuts will not break out at an unsupported area. Normally, with
a sharp cutter and a 0.010-inch finish cut, and a slow feed, chipping will
not result. High-speed steel tools with zero rake were used for turning
116 THE BELL SYSTEM TECHNICAL JOrHNAL, JANUARY 1957
Fig. 2 — Methyl methacrylate parts.
and boring operations. Standard high-speed milling cutters and end
mills were used for milling except for the cutting edges, which are honed
to a fine finish. A clearance angle of 7 degrees for milling and 10 degrees
to 15 degrees for lathe work was found moat satisfactory. In lathe work,
the general rule was light feeds (0.003 inch-0.005 inch) and small depth
of cut. However, the depth of cut could be safely varied over a wide
range depending upon many factors, such as type of part, quality of
finish, machine and tool rigidity, effective application of coolant, and
tooling to support and clamp the part. In one operation of boring a
1^-inch diameter by 4|-inch deep blind hole within ±0.002 inch, the
boring terminates in simultaneously facing the l)ottom of the hole square
with its axis. A cut ^-inch deep with a light feed was taken with a spe-
cially designed boring tool with the coolant being fed through the shank
to the cutting edge. All completely machined parts were annealed for
12 hours at 175° F.
HIGHLIGHTS IN ASSEMBLY AND BKAZING
Repeater units are encased in hardened steel rings which previously
had been tested at 10,000 pounds per square inch hydraulic pressure.
This pressure is approximately 50 per cent higher than the greatest pres-
sure expected at ocean bottom. The steel rings were encased in a copper
sheath and closed at each end with a glass-to-Kovar seal, with the cen-
tral conductor coming through the glass to the outside. The copper sheath
was then shrunk to the steel rings and glass seals using 6000 pomids per
FLEXIBLE KEPEATEH iLiXUFACTURE 117
square inch hydraulic pressure, aucl the glass seal was then high-fre-
quency brazed to the copper sheath.
To keep the ocean bottom pressure off the glass seals and also to ter-
minate the cable insulation, a rubber seal is brazed in to the copper con-
tainer tube adjacent to each glass seal. This rubber seal consists of rub-
ber bonded to l)rass, which has been brazed to the copper portion of the
seal. The rubber terminates in polyethylene through five steps of com-
pounds containing successively less rubber and more polyethylene. The
polyethylene can be readily bonded by molding to the polyethylene in-
sulation of the cable. The central conductor passes through a central
brass tube in the rubber seal, which is also bonded to the loibber.
To protect the rubber seals from the deleterious effects of salt water
immersion for long periods of lime, a copper core tube is brazed over
each rubber seal. The core tube is arranged to equalize the pressure in-
side and out when submerged at ocean bottom pressure. This is accom-
plished with a bulge of ncoprene filled with polyisobutylene, on the far
end of the core tube, which transmits the pressure to the inside of the
core tube seal.
To make doubly sure that no salt water reaches the rubber seal, a
copper cover is brazed into the container outside the core tube connector
on each end. This cover is also brazed to the core tube connector. The
interatice between each of the above four seals is filled with polyiso-
butylene, which is viscous and inert and has very good insulating
Each end of the repeater closure (Fig. 3) contains five successive brazed
joints. Any one of these ten brazes, if not perfect, could cause the loss
of the repeater closure and jeopardize the entire repeater. All of these
brazes were made with the repeater in u vertical position to insui'e an
even di.stribution of the brazing alloy fillet around the jomt.
An upending device was provided at the pit brazing location to raise
the repeater on its carrier to a ^-ertical position with either end up and
move it into position for brazing. The repeaters were brought into the
brazing area on an overhead monorail and an electric hoist. The shorter
repeater assemblies, before core tube and cable stub assembly, were up-
ended by hand and brazed from a raised platform.
It was necessary to make all of these brazes bj' high-frequency induc-
tion heating, since the heat must be intense, contained within a ^-ery
narrow band, evenly distributed, and the area protected from oxidation
by a somewhat reducing atmosphere. The heat must be very intense
since the time interval for the shortest braze was 10 seconds maximum
and the longest was 30 seconds. A large part of the heat was dissipated
FLEXIBLE REPEATER HL\NUFACTUHE 119
by being conducted at a high rate from the copper parts to the water
in the cooling jackets used to contain the heat in a very narrow band.
Circulating cooling water within a jacket prevented heat from being
conducted down the copper container tube to the preceding seals or to
the repeater unit. This water-cooled jacket was positioned only f inch
below the inductor, and the water was in intimate contact with the con-
tainer tube, which Is sealed off at both ends with rubber "0" rings. In
addition, for the glass seal braze, the glass inside the seal cavity was kept
covered with water during the heat cycle. The water was fed in and si-
phoned out to a constant level which was kept under observation by
the operator and the inspector to make sure that the glass was covered
at all times. The rubber seal was also water jacketed on the inside of
the seal to pre^'ent deleterious effects of the heat on the rubber insulation
around the central conductor. The imier cover braze was quenched before
the 10-second maximum interval had expired to insure that the heat did
not penetrate to the polyisobutylene at a sufficient rate to deteriorate
it or the rubber inside.
Distribution of the heat around the container tube at the braze area
was controlled by locating the work in the inductor so that the color
came up essentially evenly all the way around and at the proper level to
bring a fillet up to the top of the braze joint within the allowable time
limit. The time limit was determined by experiment so that none of the
previously assembled parts were damaged by the heat. This determina-
tion of the proper heat pattern and the prevention of overheating re-
quired the development of considerable skill on the part of the operator.
The variables encountered made it essential to rely on an operator to
control the heat rather than to utilize the timer with which the induction
heating equipment is normally controlled.
The area to be heated for brazing was protected from oxidation by
enclosing it in a separable tran.sparont plastic box and flooding the in-
terior with a gas consisting of 15 per cent hydrogen and 85 per cent nitro-
gen. This atmosphere is somewhat reducing and not explosive. The
brazing surfaces of the parts were chemically cleaned immediately before
assembly and extreme care was exercised to keep them clean until brazed.
The container tube was shrunk to the respective glass, rubber, core
tube, and cover seals using hydraulic pressure so that the surfaces to be
brazed and the brazing alloy were in intimate contact within the brazing
area. If the parts were clean and kept from oxidizing by the protective
atmosphere, the alloy would flow upward by capillary action and form
a fillet around the top of the seal, impervious to any leak.
The braze in each case was then leak tested with a heUum mass-spec-
120 THE BELL SYSTEM TECHNICAL JOUHNAL, JANUARY 1957
trometer type leak detector. A gas pressure of helium at least 25 per cent
greater than the maxknum pressure to be encountered at ocean bottom
was used. In addition, a radioisotope was used to test the effectiveness
of the final tabulation pinch welds and o\'ei-brazes which were kept open
for the leak tests under high pressure helium. These tests were made
with water pressure about 25 per cent greater than the maximum ocean
The completed repeater was inserted in a chamber 80 feet long; the
chamber was then filled with water and the pressure raised to 7,500
pounds per square inch and held at that pressure for at least 15 hours.
At the end of this period the closure had to show no sign of crushing or
The repeater unit sealed in the closure must be extremely dry to func-
tion properly. Any water vapor which might remain after the closure is
sealed, or enter during the estimated 20-year minimum life, must be
scavenged. A sealed desiccator with a thin diaphragm was, therefore,
assembled into the repeater unit sections. After completely drying and
seahng the repeater unit except for one tubulation, the diaphragm of
the desiccator was ruptured by dry nitrogen pressure and with the en-
closure filled wdth dry nitrogen the final tubulation was immediately
sealed off. To insure that the diaphragm was actually broken, a micro-
phone was strapped to the outside of the repeater over the location of
the desiccator and a second microphone arranged at the end of the
closure to pick up background noises. A pen recorder was used to record
the sound from the two microphones and also the change in nitrogen
pressure. Three simultaneous pips on the chart gave definite indication
that the diaphragm had ruptured and that the desiccant had been ex-
posed to the internal atmosphere of the repeater.
QUARTZ CRYSTAL UNITS MANUFACTURED AT KEARNY
The primary purpose of the crystal unit is to provide the means of
identifying and measuring the gain of each repeater in the cable. This
basic crystal design is in common usage. The exacting specifications for
this apphcation, however, imposed many problems and deviations from
normal crystal manufacturing processes.
Raw Quartz was specially selected for this crystal unit. The manu-
facturing process of reducing the quartz to the final plate followed the
recognized methods through the roughing operations. Due to the rigid
end requirements, the finishing operations were performed xmder labora-
tory conditions. Angular tolerances were one-third of normal limits. No
evidence of surface scratches, chipped edges or other surface imper-
FLEXIBLE REPEATER MANUFACTURE 121
fectioiis visible under 30X magnification were permitted. This resulted
ill a process shrinkage five times that experienced in normal crystal plate
In this use, the crystal units were required to meet performance tests
at currents as low as one-thousandth of a microampere — far below the
current \'alues usually encountered. Improved soldering techniques had
to be developed for soldering the gold plated phosphor bronze and nickel
wires used, because it was found that the electrical performance of the
units was directly related to the ([uality of soldered connections.
Although one-seventh of Western's production of quartz crystal units
iU'e in glass enclosures, the applicable techniques in glass working re-
(juired a complete revision. Glass components such as the stem and bulb
purchased from established sources were found to be far below the stand-
ard required for this crystal unit. For example, the supplier of the glass
tubing used in the manufacture of .stems was required to meet raw mate-
rial specifications that embodied coefficient of thermal expansion, soften-
ing point of glass, density, refractive index, and volume resistivity. The
glass stems made from this tubing by regular manufacturers were found
unacceptable and the processes used by these sources could not be readily
adapted to meet the desired specifications. The glass stems contained
four lead wires made from 30-mil Grade "A" nickel wire butt welded to
10-mil light borated Dumet wire. To assure the quality of the metal to
glass seal, each wire was inspected under SOX magnification for tool
marks and other surface imperfections. The finished stem assemblies
were inspected under 30X magnification for dimensions, workmanship,
cleanliness and minute glass imperfections, then individually stored in a
sealed plastic envelope.
The glass bulb in this crystal unit is known as the T921 design com-
monly used in the electron tube industry. The high quaUty required,
howe\'er, made TOO per cent inspection necessary. Examination under
;iOX magnification resulted in rejected bulbs for presence of scratches,
open bubbles, chips and stones. Physical limits for inside and outside
diameters as well as wall thickness were causes for additional rejects.
Only one per cent of the commercial bulbs were found acceptable, and
these were also stored in a sealed plastic envelope.
The final major assembly operation consisted of seahng the glass bulb
to the stem which had had the crystal sub-assembly welded to the nickel
wires. The techniques for "sealing in" used in quartz crystal or electron
tube manufacture were unsuited. Two important factors in this crystal
unit, which required the de\'elopment of new processes, were the prox-
imity of soft soldered connections to the sealing fires and the demands
122 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957
FLEXIBLE REPEATER MANUFACTURE 123
that the glass seal contain a minimum of residual tensile stress. These
two problems were resolved collectively by performing the sealing opera-
tion on a single spindle glass seaUng machine. Accm^ate positioning of
the glassware and sealing fires, together with precise timing and tem-
perature controls, achieved the desired results.
Evaluation of residual stresses were made by inspections using a
polarimeter and by a thermal shock test. The maximum safe stress was
established at 1.74 KG/mm^. The thermal shock test required successive
immersion of the unit in Ijoiling water and ice water. The electrical char-
acteristics of these units exceeded all others made previously by Western
Electric. The ratio of reactance to effective resistance ("Q") was greater
than 175,000 — twice that ever previously produced and 17 times that
required in the average filter crystal.
Stability for frequency and resistance was assured by a 28-day aging
test. During this period, precise daily resonant frequency and resistance
measurements were recorded against temperature within 0.1° C. The
maximum permissible change was 0.0005 per cent in frequency and +5
per cent to — 10 per cent in resistance.
GLASS SEALS MANUFACTURED AT ALLENTOWN
The glass seal used to close each end of the container for the repeaters
and equalizers is manufactured at the AUentowii Works of the Western
The unit is essentially a glass bead-type seal. It insulates the central
conductor of the repeater from the container and serves as a final vapor
barrier between the cable and the interior of the repeater. As such, it
backs up several other rubber and plastic barriers as shown in Fig. 3.
Fig. 4 shows the various components, subassembhes, and a cross-sec-
tion of the unit. The unit consists of the basic seal brazed in the Kovar
outer shell, to which is brazed, a copper extension provided with two
brazing-ring grooves. One of these grooves is used in brazing the seal,
along with support members, into a length of container tubing in the
same manner as the seal is ultimately brazed into the repeater. Packaging
of the seal in this manner was necessary to pressure test the seal. Under
test, in a specially constructed chamber 10,000 psi of helium gas pressure
was applied to the external areas of the packaged glass seal and a mass
spectrometer type leak detector was connected through the tubulation
to the internal cavity of the packaged unit. In this manner, the interface
of the glass to metal seal, the brazed joints, and the porosity of the metal
were checked for leakage. The unit is left in this package for delivery
to provide protection during shipment. Before the seal could be used,
124 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957
it was machined from the package by cutting the copper extension to
length, leaving the second groove for use in brazing the seal to the re-
peater and removing the container tubing and the support members.
The basic seal consists of the cup, central conductor and glass. The
cup (smaller cylindrical item in the upper lefthand corner of Kig. 4)
was machined from Kovar rod. The wall of the cup is tapered from a
thickne-ss of 0.025 inch at the base to 0.002 inch at the lip. The last 0.000
inch of the lip is further tapered from this 0.002 inc-h to a razor edge.
The internal surface is better than a 63-micro-inch turned finish and
was also liquid honed to give it a uniform matte finish. The central con-
ductor (slim piece in the upper right-hand corner of Fig. 4) was also
machined from Kovar rod. Both the cup and central conductor were
further processed by pickling, hypersonically Cleaning in deionized water,
and decarburizing. The glass, a borosilicate type of optical quality, was
cut from heavy walled tubing. The glass tubing was hand polished,
lapped and etched to remove surface scratches, and to arrive at the spe-
cified weight. It was also fire polished and hypersonically cleaned to
remove all traces of surface imperfections and to assure maximum clean-
In order to make the basic glass seal, the metal parts had to be oxidized
under precisely controlled conditions. For the oxidizing operation, a
suitable fixture was loaded mth brazed shell-cup assembhes, central
conductor assemblies, and a Kovar disc, which had been prepared in
precisely the same manner as the cups and central conductors. The
disc was carefully weighed before and after oxidizing and the increase in
weight divided by the area involved yields the weight gain due to oxida-
tion for each run. Limits of 1.5 to 2.5 milligrams per square inch of oxide
were set. This operation was performed by placing the loaded, sealed
retort, through which passed a metered flow of dried air, into a furnace
for a specified time-temperature cycle.
In the glassing operation the oxidized shell assembly, the carbon mold
and the central conductor were placed in a fixture and held in the proper
relation.ship. The carbon mold served to support the glass, while it was
being melted, in that section between the cup and central conductor
where the glass was normally unsupported. The prepared cut glass tubing
was loaded into the Kovar cup and the fixture was sealed into the retort.
During the glassing cycle, a constant flow of nitrogen passed through
the retort to provide an atmosphere which minimized any reduction or
further oxidation of the already carefully oxidized parts. After the proper
purging period, the retort was placed in the furnace. In the furnace, the
glass melted and formed a bond with the oxidized Kovar of the cup and
FLEXIBLE REPEATER MANUFACTURE 125
centi'al conductor to form the seal. After the specified temperature-time
cycle, the retort was removed from the furnace, allowed to partially
cool and then placed into an annealing oven.
Vertical furnaces and retorts were used for brazing, decarburizing,
oxidizing and glassing. By varying the type of gases flowing into the
retorts, atmospheres which are reducing, oxidizing, or neutral were ob-
tained. To provide maximum uniformity of process, separate retorts
and holding fixtures were provided for operations involving hydrogen
and for air-nitrogen operations, so that a retort or a fixture used for
hydrogen treatments was never used for oxidizing or glassing.
PILOT AN]) REGULAR PRODUCTION
We called our first efforts Practice Parts and Training; the next we
called Pilot Production. Next, certain items identified as Trial Laying
Repeaters and Oscillators were manufactured for use in "proving in"
the ship laying gear. To prove in manufacturing facilities, a few un-
equipped housings were made without the usual electrical components
normally in a repeater. Similarly, each of the apparatus components
and parts required exploratory and pilot effort before regular production
could be undertaken.
As might be expected, the manufacturing yield of components meeting
all requirements was very low during the early stages of the undertaking.
However, substantial improvement was brought about as experience
was gained. Comments on some of the production problems, highlights,
and yield results, follow.
Paper Capacitors were manufactured only after painstaking qualifying
trials and tests had been performed on each individual roll of paper.
Cycling and life testing, procurement of acceptable ceramic parts and
gold-plated tape and cans, selection and matching of rolls of paper for
winding characteristics, and similar problems, all had to be completely
resolved to a point of refinement previously unattempted for telephone
Composite percentage yield for all operations on paper capacitois is
shown in Fig. 5. Yield is shown as the ratio of finished units of acceptable
quality to the number of units started in manufacture.
Afira Capacitors were made from only the most meticulouslj' selected
laminations, as mentioned earlier. Even the best mica is particularly
susceptible to damage in processing. In spite of experience and knowl-
edge of this, the multiple handling of the laminations contributed an
unusually high material shrinkage as each separate lamination needed
to be cleaned, then handled individually many times through the proc-
126 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957
1 1 1 1 1 1 1 1 1 1 1
1954 1955 1956
Fig. 5 ^ Paper capacitor yield.
esses. The art of silk screening was applied to deposit silver paste in a
specific area or areas on each side of a lamination. A sharply defined
rectangular area was required so that when superimposed one over an-
other the desired capacitance would be obtained. Cementing of mica
laminations onto machined methacrylate forms presented some addi-
tional problems through the bowing of the mica laminations as the ce-
ment cured. Obtaining screens that would give the proper length and
width dimensions for the coated area, was another problem. A silk screen
woven of strands of silk obviously hniits, by the diameter of the threads,
the extent to which the dimensions of an opening may be increased or
decreased. Beryllium copper U-shaped terminals were used to clamp
the layers of mica together into a stack. Control of the pressure used in
crimping these terminals was found to be very critical in view of the
exceptionally tight limits on capacitance and stability. Fig. 6 shows the
composite yield at various times for all mica capacitors.
Resistors, There were three designs of ceramic resistors, which were
resistance -wire wound on ceramic spools. These were intended to be
assembled into the hole inside the core tube on which the paper capaci-
tors were wound. Special winding machines equipped with binocular
FLEXIBLE REPEATER MANUFACTURE
I I I I I [ I I I I I I I I I I I I I [ I I I
1954 1955 1956
Fig. 6 — Mica capacitor yield.
attachments were necessary to wind these resistors. Other resistors were
hand wound on methyl-methacrj'late forms, or on the outside of the
ceramic containers, for certain types of paper capacitors. Rough adjust-
ments were required of the lengths of resistance wire prior to winding,
and close adjustments to resistance values were mude after the ^\'indings
were completed and before leads were attached to resistors. Again it was
necessary to provide periodic samples that could be placed on life test
by the Laboratories to ascertain that the manufacturing processes were
under control. The.se .samples, in all po.ssible cases, were taken from prod-
uct that would normally be rejected becau.se of some minor defect, but
which would not in any way detract from the validity of the life tests.
The making of hard solder splices between nichrome resistance wire
and gold-plated copper leads, and keeping ceramic parts from coming
in contact with metal surfaces and thereby being contaminated because
of the ceramic's abrasive characteristics, were two major problems on
resistors. Fig. 7 indicates I'esistor yields.
Inductors comprised 20 different designs, most of which were air core,
but there were some for which it was necessary to cement permalloy
dust cores into pockets of the methacrylate form, and thereafter using
THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957
1954 1955 1956
Fig. 7 — Resistor yield.
wire on a shuttle, wind by hand the turns rciiuired to produce an in-
ductor. The.se varied from a very small inductor, smaller in diameter
than a pencil, to a fairly large "figure eight" inductor with turns having
a major diameter of about Ij inches. Each layer of a winding was in-
spected with a microscope to insure that the wire had not been twisted
or kinked, or that the insulation was damaged or uneven. Some of the
shuttles became fairly long so that they could hold the amount of wire
required to make a continuous winding. The operator's handling of this
shuttle, as she moved it down around the openings in the methacrylate
part, or placed it on a bench to proceed with the interlea^'ing tape, de-
manded considerable dexterity and concentration to insure that the
shuttle was not turned over — which in effect would put a twist in the
wire. Although best known means were used to sort cores for their mag-
netic properties prior to the time a winding was made, the limits on the
inductors themselves were so close that subsequently a large number of
windings were lost. The best cores that could be selected, plus the best
winding practice, could not produce 100 per cent of the inductors within
the required limits. Crazing of the insulation on the wire; cementing
together of two methacrylate parts or of permalloy cores into pockets of
FLEXIBLE REPEATER ULANUFACTUHE
I r \l Xy
1 1 1 1 1 1 f 1 1 1 1
1 1 1 1 11
1 1 1 1 1
1954 1955 1956
Fig. 8 — Iiidiiclor yield.
inethaciylate parts, and handling those inductors having long dehcate
leads, were the most troublesome items on this apparatus. Fig. 8 shows
manufacturing yield for inductors.
Networks combined several codes of component apparatus, such as
a mica and a paper capacitor, resistor and an inductor. Six networks
were used in each repeater unit consisting of two interstage networks,
an input, an output, and two beta networks. They demanded a most
deUcate wiring job in that stranded gold-plated copper wires had to be
joined in a small pocket in methyl methacrylate, where a minimum
amount of heat can be applied; otherwise the methacrylate is affected.
After soldering, a minimum amount of mo"\'ement of the stranded wire
was permitted, inasmuch as the soldered gold-plated copper wire be-
comes quite brittle.
Repeater Units, are wired assemblies consisting of seventeen sections
in which there are six networks, three electron tubes, one gas tube, one
crystal, three high voltage capacitors, one dessicator and two terminal
sections. The successive build-up of these materials left little chance to
make a repair because a splice in a lead was not permissible. It is during
this assembly stage that a repeater received its individual identity be-
130 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957
cause of the frequency of the particular crystal assembled into the unit.
A manufacturing yield of 100 per cent was achieved in the assembly and
wiring of repeater units.
It was necessary to calibrate the test equipment for this job very
closely. Bell Telephone Laboratories and Western Electric worked at
length to calibrate the testing details and the test sets for individual net-
works. Adjustments in components apparatus to bring the network to
the fine tolerances required were accomphshed by minute scraping of
the silvered mica on a mica capacitor or removing turns from wire-
womid inductors. The cementing of methacrylatc parts, which was a
troublesome item on mica capacitors and inductors, also had to be con-
tended with on networks.
PACKING AND SHIPPING COORDINATION
Repeaters were packed in Western Electric specially designed 34-foot
long aluminum containers, weighing 1,000 pounds. Forty of these con-
tainers were made by an outside firm. Fig. 9 shows two containers tied
down in a truck trailer. The repeaters were nested in a pocket of poly-
ethylene bags containing shaped rubberized hair sections in order to
cushion the repeaters during their subsequent handhng and transporta-
tion. The instrumentation required with each case was tested, properly
set, and inspected prior to its use on each outgoing case. The instruments
were a shock recorder to register shocks in three planes, and a thermome-
ter to register the minimum and maximum temperatures to which the
repeater had been exposed. Arrangements were made with a commercial
trucking company to provide three specially equipped truck trailers,
which could be cooled by dry ice during hot weather and warmed by
burning bottled gas during cold weather so as to control temperature
within the 20-degree F. to 120-degree F. called for in the repeater spe-
Appointment of a shipping coordinator supervisor added tremendously
to the smooth functioning of services and provided the continuing vigi-
lance required to protect repeaters and deliver them to the right place
at the right time. His responsibihty was to coordinate all the shipping
information and arrangements from the time the item was ready for
packing at the Hillside plant, through all trucking arrangements to the
armoring factory, to the airport, to England, and to follow, with sta-
tistical data and reports, each enclosure until we were able to record
the date on which the repeater was laid or stored in a depot.
FLEXIBLE REPEATER MANUFACTURE
Fig. 9 — Shipping contaiuera.
INSPECTION PLAN AND PHOCEDUKES
It is axiomatic that quality is not obtained by inspection but must be
built into the product. However, the Inspection Organization docs have
the responsibihty of certifying that the desired quaUty exists. Our eval-
uation indicated that the ordinary inspection "screening" would be
inadequate to insure the high degree of integrity demanded and that
additional safeguards would have to be provided. These controls were
achieved, in a practical way, by:
(1) Selective placement, intensive training and subsequent qualifica-
tion testing of all persoimel.
(2) Inspection during manufacturing operations in addition to in-
132 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957
spection of product after completion, and regulating inspection so that
critical characteristics received repetitive examination during the process
of manufacture and assembly.
(3) A maintenance program for inspection and testing facihties which
provided checks at considerably shorter intervals than is considered
(4) Inspection and operating records and reports that point out areas
for corrective measures.
(5) Records of quality accuracy for all inspection personnel as an aid
in maintaining the high quality level.
(6) Verification of all data covering process and final inspection as a
certification of the accuracy of these data and that the apparatus satis-
factoiily meets all requirements.
Selection and Training of Inspection Personnel ^.
The quality of a product naturally depends upon the skills, attitude,
and integrity of the personnel making and inspecting it. It was realized
that in order to develop the high degree of efficiency in the inspection
organization necessary to insure the integrity of the product, personnel
of very high caliber would be required. These employees would have to
be (1) experienced in similar or comparable work, (2) they would have
to be precise, accurate and, above all, dependable, (3) in order to reduce
the possibihty of contamination and damage they would have to be neat
and careful, and (4) they would require the ability to work in harmony
with other employees, often as a member of a "team," in an environ-
ment where their work would be under constant scrutiny.
Most of the inspection employees selected to work at Hillside were
transferred from the Kearny Plant and had an average Western Electric
sei-vice of twelve years. They were hand-picked for the attributes out-
lined above, and the "screening" was performed by supervision through
personal interviews supplemented by occupational tests given by the
personnel department. These tests, which are in general use, are designed
to evaluate background and physical characteristics, and they were
given regardless of whether the employee had or had not previously
The following group of tests is an example of those given inspectors
and testers of apparatus components :
(1) Electrical — ac-dc theory and application.
(2) Ortho-Rater — Eye test for phoria, acuity, depth, and color.
(3) Finger Dexterity — Ability and ease of handUng small parts.
FLEXIBLE REPEATER MANUFACTURE 133
(4) Special — Legibility of handwriting, ability to transcribe data
and to use algebraic formulae in data computations.
The general plan of visual and mechanical inspection consisted of:
(1) Inspection of every operation performed — and in many cases
partial operations — during the course of manufacture. This is of par-
ticular importance where the quality characteristics are hidden or inac-
cessible after completion of the operation.
(2) Repeated inspection at subsequent points for omissions, damage
(3) Rejection of product at any point where there was failure to ob-
tain inspection or where the results of such inspection had not been
Most of the visual inspection was performed at the operators' posi-
tions to reduce, to a minimum, the amount of handling that could result
in damage and contamination.
Visual inspection covered three general categories:
(1) Inspection of work after some or all operations had been com-
pleted, such as the machining of parts.
(2) Inspection at those points where successive operations would cover
up the work already perform^ed. An example of this is the hand winding
of toroidal inductors where each layer of wire was examined under a mi-
croscope for such defects as twists, cracks, and crazes in enamel insula-
tion, spacing and overlapping of turns, and contamination before the op-
erator was allowed to proceed with another layer. While being inspected,
the work remained in the holding fixture, which was hinged in such a man-
ner as to permit inspection of both top and bottom of the coil. Inductors
received an average of 13 and a maximum of 2G visual inspections during
(3) Continuous "over-the-shoulder" inspection, where strict adher-
ence to a process was required or where it was impossible to determine,
by subsequent inspection, whether or not specific operations had been
performed. In these cases, the inspector checked the setup and facilities,
observed to see that the manufacturing layouts were being followed,
that the operations were being performed satisfactorily, and that spe-
cifications were being met.
The electrical testing, in itself, was not unusual for carrier apparatus
and runs the gamut from dc resistance through capacitance, inductance,
134 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957
and effective resistance, to transmission characteristics in the frequency
band 20-174 kc. What was unusual were the extremely narrow hmits
imposed and the number and variety of tests involved as compared to
those usually specified for commercial counterparts.
The following two examples will serve to illustrate the extreme meas-
ures taken to prove the integrity of the product :
(A) One type of Resistor was wound with No. 46 mandrelated nichrome
wire to a value of 100,000 ohms plus or minus 0.3 per cent. This resistor
received six checks for dc resistance, five for instantaneous stability of
resistance and two for distributed capacitance, at various steps in the
process which included six days' temperature cycling for mechanical
stabilization. This resistor was considered satisfactory, after final anal-
ysis of the test results, if: (a) The difference in any two of the six resist-
ance readings did not exceed 0.25 per cent, (b) The change in resistance
during cycling was not greater than 0.02 per cent, (c) The "instantaneous
stability" (maximum change during 30 seconds) did not vary more than
0.01 per cent. In addition, it was required that the distributed capaci-
tance, minimum 7, maximum 10 mmf, should not differ from any other
resistor by more than 2 mmf.
(B) For high voltage paper capacitors, the 0.004-inch thick Kraft
paper, which constitutes the dielectric, was selected from the most
promising mill lots which the manufacturers had to offer. This selection
was based on the results obtained from tests that involve examination
for porosity, conducting material and conductivity of water extractions.
These tests were followed by the winding and impregnation in Halowax
of test capacitors. The test capacitors were then subjected to a direct
voltage endurance test at 266 degrees F for 24 hours.
Samples of prospective lots of paper, which have passed the above
test, were then used to wind another group of test capacitors that were
subsequently impregnated with Aroclor and sealed. 1,500-volt dc was
then apphed to the capacitors at 203° F for 500 hours. In case of failure,
a second sampling was permitted.
After the foregoing tests had been passed, the supplier providing the
particular mill lot was authorized to sht the paper. Upon receipt, six
special capacitors were wound, using a group of six rolls of the paper
being qualified. These capacitors were then impregnated, checked for
dielectric strength at 3,000-volt dc, and measured for capacitance and
insulation resistance. The capacitors were then given an accelerated life
test at 2,000-volt dc, temperature 150° F, for 25 days. Each lot of six
satisfactory test capacitors qualified six rolls of paper for use.
Product capacitors were then wound from approved paper, and the
dry units checked for dielectric strength at 300-volt dc. Capacitance
FLEXIBLE REPEATER MAISfUFACTURE 135
was checked and units were then assembled uito cans and ceramic covers
soldered in place. Assemblies were pressmized with air, through a hole
provided for the purpose, while the assembly was immersed in hot water
to determine if leaks were present. Capacitors were then baked, vacuuni
dried, impregnated, pressurized with nitrogen, and sealed oJT. The com-
pletely sealed units were then placed in a vacuum chamber at a tem-
perature of 150° F, 2 mm. mercury, for 3 hours to check for oil leaks.
Capacitance was I'echecked and insulation resistance measured.
After seven days, capacitors were unsealed to replenish the nitrogen
that had been absorbed by the oil, resealed and again vacuum leak tested.
An X-ray examination was then made of each individual unit to verify
internal mechanical conditions. Capacitors were then placed in a tem-
perature chamber and given the following treatment for one cycle:
16 hours at 150°F; 8 hours at 75°F; 16 hours at O^F; 8 hours at 75°F.
At the end of ten days, or 5 cycles, the insulation resistance and con-
ductance was measured and a norm established for capacitance.
Capacitors were then recycled for ten days, and, if the capacitance
had not changed more than 0.1 per cent, they were satisfactory to place
on production life test. If the foregoing conditions had not been met,
the capacitors were recycled for periods of ten days until stabilized.
At that time, 10 per cent of the capacitors in every production lot
were placed on "Sampling Life Test", which consisted of applying 4,000-
volt dc in a temperature of 150°F for 25 days. At the same time, the
balance of the capacitors in the lot were placed on production life test
at 8,000-volt dc in a temperature of 42°F for 26 weeks. At the end of
this time, the insulation resistance w^as measured and the capacitance
checked at To^F and at 39°F. The difference in capacitance at the two
temperatures could not exceed +0.001, —0.005 mf, and the total ca-
pacitance could not exceed maximum 0.3726, minimum 0.3674 mf. The
capacitance from start to finish of the life test could not have changed
more than plus or minus 0.1 per cent.
If all of the preceding requirements had been satisfied, the particular
lot of capacitors described was considered satisfactory for use.
The foregoing examples are typical of the procedures evolved for
insuring, to the greatest degree possible, the long, trouble-free life of all
apparatus used in the repeater.
There were many new and involved tests which were developed and
applied to the manufacture of repeaters. One of the most unique is the
use of a radioisotope for the detection of leaks imder hydrauHc pressure.
136 THE BELL SYSTEM TECHNICAL JOURNAL, JANXJAnY 1957
The initial closure operations consisted of brazing into each end of
the repeater housing a Kovar-to-glass seal. These seals are equipped
with small diameter nickel tubulations which were used to flush and
pressurize the repeaters with nitrogen. After these operations had been
performed, one of the tubulations was pinchw^elded, overbrazed and
coiled down into the seal cavity. The repeater was then placed in a pres-
sure cylinder with the open tubulation extending through and sealed to
the test cylinder. A mass spectrometer was then attached to the tubula-
tion and the te.st cylinder pressurized with helium at 10,000 psi. At the
conclusion of this test the repeater was removed from the test cylinder
and, after breaking the desiccator diaphragm, the remaining open tubu-
lation was pinehwelded and overbrazed. At this point, it became neces-
sary to determine whether the final phichweld and overbrazing would
leak under pressure.
Since there was no longer any means of access to the inside of the
repeater, aU testing had to be done from the outside. This was accom-
plisheil by filhng the glass seal with a solution of radioisotope cesium
134, which was retained by a fixture. The repeater was then placed in a
test cylinder and hydrauhc pressure applied, which was transmitted to
the radioisotope in the fixture. After 60 hours under pressure, the re-
peater was removed from the cylinder and the seal drained and washed.
An examination was then made with a Geiger counter to determine if
any of the isotope had entered the final weld.
The washing procedure, after application of the isotope solution, in-
volved some sixty operations with precise timing. In the case of the re-
peater at the rubber seal stage where both ends were tested, it was
desirable that these operations be performed concurrently. This was
accomphshed by recording the entire process on magnetic tape which,
when played back, furnished detailed instructions and exact timing.
RAW MATERLAL INSPECTION
As might be expected, raw materials used in the project were very
carefully examined and nothing left to chance. Every individual bar,
rod, sheet, tube, bottle or can of materials was given a serial number and
a sample taken from each and similarly identified. Each sample was
then given a complete chemical and physical analysis before each corre-
sponding piece of material was certified and released for processing. In
many cases, the cost of inspection far exceeded the cost of the material.
However, the discrepancies revealed and the assurance provided, more
than justify the expense.
Detailed records of all raw material inspection were compiled and
furnished to the responsible raw material engineer who examined them.
FLEXIBLE REPEATER fcL\NUFACTUHE 137
critically, as an additional precaution before the material was released
to the shop.
To eUminate, as much as possible, the human element in providing
assurance that all prescribed operations had been performed satisfac-
torily, inspected properly and the resuhs recorded, means were estab-
lished to compile a complete history of the product concurrent with
manufacture. This was accomplished through the provision of permanent
data books of semilooseleaf design, which require a special machine for
removing or inserting pages.
Each of these books covered a portion of the work involved in pro-
ducing a piece of apparatus and contained a sequential list of pertinent
operations and requirements prescribed in the manufacturing process
specifications. Space was provided, adjacent to the recorded information,
for both the operator and inspector to affix their initials and the data.
A reference page in the front of each book identified the initials with
the employees' names. All apparatus was serially numbered and the
data were identified accordingly. If a unit was rejected, that serial num-
ber was not reused.
These data books, in addition to estabUshing a complete record of
nianufacture, provided a definite psychological advantage in that people
were naturally more attentive to their work when required to sign for
As pointed out previously, every precaution was exercised in selecting
and training inspection personnel assigned to the project. However, it
was realized at the outset that human beings are not infallible and that
insurance, to the greatest degree possilile, would ha\'e to be provided
against the probability of errors in observation and jugment. Quahty
accuracy evaluation procedures were, therefore, established for deter-
mining the accui'acy of each inspector's performance.
Quality accuracy checking was performed by a staff of five Inspection
Representatives and involved an examination of the work performed
by inspectors to determine how accurately it was inspected. Materials
which the inspector accepted and those which had been rejected were
VERIFICATION AND SUMMARY OF DATA
As an added measure of assurance as to the integrity of the product,
procedures were established for verifying and simimarizing the inspec-
tion records for each serially numbered component, up to an including
138 THE BELL SYSTEM TECHNICAL JOURNAL, JANUARY 1957
Verification involved a complete audit of the inspection records to
provide assurance that all process operations were recorded as having
been performed satisfactorily, that the prescribed inspections had been
made, and that the recorded results indicated that the product met all
of the specified requirements. This work was performed by a group of
six Inspection Eepresentatives who had considerably experience in all
phases of inspection and inspection records.
As the verification of a particular piece of apparatus proceeded, a
verification report was prepared which, when completed, contained the
most pertinent inspection data, such as:
(1) Recorded measurements of electrical parameters.
(2) Values calculated from measurements to determine conformance.
(3) Confirmation that all process and inspection operations had been
(4) Identification (code numbers and serial or lot numbers) of mate-
rials and components entering into the product at each stage of manu-
The verification report usually hsted the data for twenty serial num-
bers of a particular code of apparatus along with the specified require-
ments. Included, also, was a cross-reference to all the inspection data
books involved so that the original data could be located easily. These
verification reports were prepared for all apparatus up to and including
the finally assembled and tested repeaters.
The following gives an indication of the number of items examined
in the verification of one complete repeater:
Items verified in data books 17 , 593
Items verified on recorder charts 1 , 142
Calculations verified 1 , 580
Number of entries on verification reports 4 , 070
Verification reports, in addition to presenting the pertinent recorded
data, provided a "field" of twenty sets of measurements from which it
was easily possible to spot a questionable variation. For example, it was
the adopted practice on this project to examine, critically, any charac-
teristic of a piece of apparatus, in a universe of twenty, which varied
considerably from the rest, despite the fact that it was still within limits.
While the number of cases turned up in the verification process which
have resulted in rejection of product are relatively few, we believe that
the added insurance provided, and the psychological value obtained,
considerably outweigh the cost.